Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Efficient and Versatile Quadrupedal Skating: Optimal Co-design via Reinforcement Learning and Bayesian Optimization

Hanwen Wang, Zhenlong Fang, Josiah Hanna, Xiaobin Xiong

Abstract

In this paper, we present a hardware-control co-design approach that enables efficient and versatile roller skating on quadrupedal robots equipped with passive wheels. Passive-wheel skating reduces leg inertia and improves energy efficiency, particularly at high speeds. However, the absence of direct wheel actuation tightly couples mechanical design and control. To unlock the full potential of this modality, we formulate a bilevel optimization framework: an upper-level Bayesian Optimization searches the mechanical design space, while a lower-level Reinforcement Learning trains a motor control policy for each candidate design. The resulting design-policy pairs not only outperform human-engineered baselines, but also exhibit versatile behaviors such as hockey stop (rapid braking by turning sideways to maximize friction) and self-aligning motion (automatic reorientation to improve energy efficiency in the direction of travel), offering the first system-level study of dynamic skating motion on quadrupedal robots.

Efficient and Versatile Quadrupedal Skating: Optimal Co-design via Reinforcement Learning and Bayesian Optimization

Abstract

In this paper, we present a hardware-control co-design approach that enables efficient and versatile roller skating on quadrupedal robots equipped with passive wheels. Passive-wheel skating reduces leg inertia and improves energy efficiency, particularly at high speeds. However, the absence of direct wheel actuation tightly couples mechanical design and control. To unlock the full potential of this modality, we formulate a bilevel optimization framework: an upper-level Bayesian Optimization searches the mechanical design space, while a lower-level Reinforcement Learning trains a motor control policy for each candidate design. The resulting design-policy pairs not only outperform human-engineered baselines, but also exhibit versatile behaviors such as hockey stop (rapid braking by turning sideways to maximize friction) and self-aligning motion (automatic reorientation to improve energy efficiency in the direction of travel), offering the first system-level study of dynamic skating motion on quadrupedal robots.
Paper Structure (17 sections, 9 equations, 9 figures, 2 tables)

This paper contains 17 sections, 9 equations, 9 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Quadrupedal Locomotion
  4. Robot Skating with Passive Wheels
  5. Robot Hardware and Control Policy Co-Design
  6. Quadrupedal Skating via Co-Design
  7. Co-Design via Bilevel Optimization
  8. Mechanical Design of Roller Support
  9. Algorithmic Implementation
  10. Inner Loop Control Policy Optimization via RL
  11. Outer Loop Design Optimization via BO
  12. Results and Analysis
  13. Baseline via Human-Engineered Roller Skating
  14. Necessity of Co-Design for Energy Efficiency
  15. Comparison between Base Frame Command and World Frame Command
  16. ...and 2 more sections

Figures (9)

  • Figure 1: Quadrupedal skating setup with passive wheels. Each foot is equipped with a 3D-printed roller support that holds a passive wheel. In the default standing configuration, the wheel yaw installation angle $\psi$ is defined as the deviation of the wheel’s x-axis, $x_\text{wheel}$, from the x-axis of the robot’s sagittal plane. This angle is our key design parameter.
  • Figure 2: Bilevel co-design framework for quadrupedal skating. The outer loop uses BO to propose the next candidate design $\mathbf{d}_{\text{next}}$. For each candidate design, the inner loop uses RL to optimize the policy $\pi_\theta$. The resulting performance $J(\pi_\theta^*, \mathbf{d}, \mathcal{T})$ is fed back to BO, enabling efficient search over design-policy pairs that maximize skating performance.
  • Figure 3: Naive P configuration ($\psi = 0^\circ$) on Unitree Go1 UnitreeGo1 yields uncontrollable forward velocity $v_x$ due to leg kinematic constraints. Similar limitations arise on many state-of-the-art quadrupedal robots.
  • Figure 4: Illustration of World Frame Command angular velocity tracking reward scaling to prioritize linear velocity tracking.
  • Figure 5: Illustration of the human-engineered roller wheel angles and the resultant gait pattern.
  • ...and 4 more figures